Skip to main content

Sindri Infrastructure

An overview of the design and architecture of Sindri and the advantages it brings to you as a user.

A conceptual diagram depicting interactions with Sindri's API.

Sindri provides a RESTful API designed for efficiency and scalability. It is tailored for developers who need robust solutions, in production and development, for handling zero-knowledge proofs across proving frameworks. Today, Sindri’s infrastructure is seeded with a market-leading GPU catalog to provide high availability across all our user’s proving environments across a variety of GPU types.


Sindri’s backend scales dynamically, ensuring seamless adaptation to user workloads. With smart queuing logic, it achieves efficient load balancing. Whether using dedicated or shared instances, Sindri maintains optimal response times. Users can confidently submit API calls across all circuits without any compromise in performance. This adaptability allows Sindri to effectively handle both high-volume batch and concurrent proofs, scaling effortlessly with the evolving needs of users' environments.

A sketch of responsive autoscaling, with a maximum limit of three workers.


Sindri streamlines the resource management process with its self-service provisioning. This feature empowers users to customize resources to fit the specific requirements of diverse circuits and environments. The platform's flexible workload routing across elastic GPU and CPU backends enables you to utilize compute and infrastructure resources on-demand, optimizing for both efficiency and overhead. Consequently, Sindri serves as an invaluable tooling companion, supporting your journey from development to production deployments.

Cost Effective

Sindri’s serverless-style consumption model lets you pay for the compute you actually use. This aligns costs directly with usage and improves accessibility to critical infrastructure without the soft costs (uptime monitoring, maintenance, versioning) and hard costs associated with always-on instances.


Sindri intelligently manages storage for zero-knowledge proofs and their associated public data. While the public data of a proof is persistent within our storage, the private inputs are not saved. Moreover, you can (and should) send encrypted private inputs to the proof endpoint via TLS.

Our commitment extends beyond scalability solutions as we are deeply invested in the future of private computation. We aim to enhance the protection of shielded input data while maintaining the simplicity and ease-of-use characteristics of our API experience.


End-to-end performance is facilitated through Sindri acceleration and latency optimizations. Sindri's novel MSM algorithm, Sagittal MSM, is fully streamlined and deterministic and it is designed to be the most performant on GPUs. This means you can benefit from Sindri’s proof acceleration across a broad spectrum of widely-available hardware. Furthermore, dynamic instance configuration, memory caching, programmatic redundancy, and data proximity measures push down latency overhead.